On Mon, Jan 19, 2015 at 10:37:58AM -0500, Benjamin Tissoires wrote:

Hi Antonio,

[adding Mika in CC, he implemented most of the ACPI and GPIO for
i2c-hid]

On Jan 17 2015 or thereabouts, Antonio Ospite wrote:

quoted

Hi,

I am trying to make the Goodix driver (drivers/input/touchscreen/goodix.c)
working with a Teclast X98 Air 3G, a tablet based on Intel Bay Trail,
but I am new to ACPI and I could use some help.

I am working with a 3.19-rc4 kernel compiled for x86_64.

This is the DSDT section in the UEFI firmware:

            Device (TCS0)
            {
                Name (_ADR, Zero)  // _ADR: Address
                Name (_HID, "GODX0911")  // _HID: Hardware ID
                Name (_CID, "PNP0C50" /* HID Protocol Device (I2C bus) */)  // _CID: Compatible ID

urgh, this is bad. It declares itself as i2c-hid, but it is not :(
Anyway, according to your logs, i2c-hid probe() just fails, so it's not
a big problem.

Actually, I think this device should use i2c-hid. All the ACPI plumbing
is there including _DSM.

What makes you think it should use the goodix driver?

quoted

                Name (_S0W, Zero)  // _S0W: S0 Device Wake State
                Name (_DEP, Package (0x02)  // _DEP: Dependencies
                {
                    GPO1, 
                    I2C5
                })
                Method (_PS3, 0, Serialized)  // _PS3: Power State 3
                {
                    If ((^^^I2C5.PMIC.AVBG == One)) {}
                }

                Method (_PS0, 0, Serialized)  // _PS0: Power State 0
                {
                    If ((^^^GPO1.AVBL == One))
                    {
                        ^^^GPO1.TCD3 = Zero
                    }

                    Sleep (0x05)
                    If ((^^^I2C5.PMIC.AVBG == One)) {}
                    Sleep (0x1E)
                    If ((^^^GPO1.AVBL == One))
                    {
                        ^^^GPO1.TCD3 = One
                    }

                    Sleep (0x78)
                }

                Method (_CRS, 0, NotSerialized)  // _CRS: Current Resource Settings
                {
                    Name (RBUF, ResourceTemplate ()
                    {
                        I2cSerialBus (0x0014, ControllerInitiated, 0x0019F0A0,
                            AddressingMode7Bit, "\\_SB.I2C4",
                            0x00, ResourceConsumer, ,
                            )
                        GpioIo (Exclusive, PullDefault, 0x0000, 0x0000, IoRestrictionInputOnly,
                            "\\_SB.GPO2", 0x00, ResourceConsumer, ,
                            )
                            {   // Pin list
                                0x0044
                            }
                    })
                    Name (ABUF, ResourceTemplate ()
                    {
                        I2cSerialBus (0x0014, ControllerInitiated, 0x0019F0A0,
                            AddressingMode7Bit, "\\_SB.I2C4",
                            0x00, ResourceConsumer, ,
                            )
                        GpioIo (Exclusive, PullDefault, 0x0000, 0x0000, IoRestrictionInputOnly,
                            "\\_SB.GPO2", 0x00, ResourceConsumer, ,
                            )
                            {   // Pin list
                                0x0044
                            }
                        GpioIo (Exclusive, PullDefault, 0x0000, 0x0000, IoRestrictionOutputOnly,
                            "\\_SB.GPO1", 0x00, ResourceConsumer, ,
                            )
                            {   // Pin list
                                0x001A
                            }
                    })

It looks like the GPIOs are correctly declared. The ACPI code should set
the client->irq auto-magically. It's not the case, so I guess Mika
should be able to tell us more on that.

The current i2c-hid.c does not cope with GPIO interrupts. I've attached
an experimental patch that should convert the driver to use them.

Antonio, can you try that out and check if i2c-hid driver gets you
working touch screen? Also please add "i2c_hid.debug=1" to the kernel
command line so we can see if it returns proper HID descriptor.